SEE IT IN ACTION
Watch it work — real terminal, real output.
No mockups. Pick who you are, then choose a scenario.
gpusmarket rent --model gemma3:4b && gpusmarket chatgpusmarket cheapest --model gemma3:4bgpusmarket rentals && gpusmarket stop <id>gpusmarket pool create && gpusmarket pool addclaude mcp add gpusmarket -- gpusmarket mcp-servegpusmarket host setupnvidia-smi && gpusmarket host setupdocker compose up -dgpusmarket swarm join && gpusmarket swarm rentFor AI developers
Stop overpaying for GPU compute.
Same GPUs. Same models. 80% cheaper than cloud providers. No setup, no lock-in.
80% cheaper
The same RTX 4090 costs $0.35/hr here vs $2.49/hr on AWS. Peer-to-peer means no cloud markup.
Per-second billing
Use a GPU for 47 seconds? Pay for 47 seconds. No hourly minimums, no idle charges, no commitments.
Instant API endpoint
Click rent, get an API key. No Docker, no SSH keys, no CUDA drivers. Just curl and go.
Drop-in replacement
Standard OpenAI SDK. Change one line of code. Works with every tool you already use.
Works with your stack
How it works for GPU owners
Install AI software
Download Ollama (free, 1-click install) and pick which AI models to run. Takes 2 minutes.
List your GPU
Sign up, tell us your GPU model, and set your hourly price. We'll guide you through every step.
Get paid
People rent your GPU, money goes straight to your bank via Stripe. We handle billing, payments, and security. You just keep it running.
GPU pricing
Real market rates - updated April 2026
Your GPU earns nothing when it's idle. List it and it pays for itself.
| GPU | You earn/hr | 8 hrs idle/day | Monthly |
|---|---|---|---|
| RTX 3090 24 GB | |||
| RTX 4090 24 GB | |||
| RTX 5090 32 GB | |||
| A100 80GB 80 GB | |||
| H100 SXM 80 GB | |||
| H200 141 GB |
Save 60–85% vs AWS, GCP, and Azure. Real on-demand rates compared.
| GPU | GPUs Market | AWS / GCP | RunPod | You save |
|---|---|---|---|---|
| RTX 4090 24 GB | 34% | |||
| RTX 5090 32 GB | 44% | |||
| A100 80GB 80 GB | 65% | |||
| H100 SXM 80 GB | 80% | |||
| H200 141 GB | 73% | |||
| B200 180 GB | 72% |
Sources: getdeploying.com, runpod.io, vast.ai, aws.amazon.com - April 2026 on-demand rates. Savings vs AWS/GCP.
Why people trust GPUsMarket
You're in control
Set your own price. Choose what access to offer. Turn your listing on or off anytime.
Easy to set up
Install one free app, sign up, and follow our step-by-step wizard. No technical expertise needed.
Money goes to your bank
We take 10%, you keep 90%. Paid directly to your bank account via Stripe. No invoicing needed.
Your machine is safe
Renters never see your IP address. All connections go through our secure relay. We monitor everything 24/7.
GPU marketplace
44 GPU models · 11 available nowNVIDIA GeForce
NVIDIA Workstation
Datacenter
Apple Silicon, AMD & Intel
Need a GPU for AI? Rent one instantly.
Browse available GPUs from real people, click Rent, and start running AI models immediately. Pay only for the time you use - by the second, not the hour. Up to 85% cheaper than AWS or Google Cloud.
Browse available GPUsWhat can you do with a rented GPU?
Whether you're building a chatbot, training a model, or just experimenting - there's an option for you.
Chat with AI models
Run AI models like Llama, Gemma, or Qwen. Get an instant API you can connect to any app. No setup needed.
Browse AI GPUsTrain & fine-tune models
Get a private workspace with GPU access. Pre-built environments for PyTorch, Jupyter, and more. Upload your data and train.
Browse environmentsFull remote access
Get full access to a GPU machine. Install your own tools, run custom code, do whatever you need. Like having your own server.
Browse full-access GPUsAgentic Memory
Your LLMs remember context across sessions. Conversations auto-captured, history auto-injected. Free starter plan.
Enable MemoryDeveloper tools
Search, rent, and manage GPUs from your terminal or code.
REST API
Full REST API - search, rent, manage programmatically. Public endpoints, no auth for reads.
View API reference →Templates
Pre-built stacks: PyTorch, vLLM, ComfyUI, text-gen-webui, Jupyter. Pick and launch in 60 seconds.
Browse templates →Chat with a live LLM — or generate an image — on a community GPU swarm. Free, no signup.
Type a prompt and watch it route across a community mesh of GPUs in real time. Or hit the image tab and get a picture back in ~2 seconds. No account, no API key, no card. Then add your own GPU in one command and earn a cut every time it serves.
AI image generation, priced per image.
The same marketplace, now for pictures. Real gaming GPUs running SDXL-Turbo return an image in about 2 seconds — and you only pay for the images you actually generate.
Generate images
Standard OpenAI Images API — if your code can call DALL·E, it can call us. No queues, no credits system, no subscription. Rent an image GPU, point your app at one endpoint, pay per image, stop whenever.
POST /v1/images/generations
{ "prompt": "a red bicycle on a beach",
"n": 1, "size": "512x512" }Earn with your gaming card
An RTX 3090 or 4090 is too small for the big LLMs — but it's the perfect image-generation machine. Our ready-made SDXL-Turbo template turns your card into a paid image API: copy the stack, list it, set your price per image.
- Ready-made template — no ML setup, weights download themselves
- Your machine stays behind your firewall — our tunnel handles traffic
- You set the price per image and keep 90%
Per-second billing. No hourly rounding.
Billed per second, not rounded to the nearest hour. Stop anytime - you only pay for the exact time you used.
GPU Pools
Rent 10 GPUs. Get 1 endpoint. 10x the throughput.
10x throughput
Round-robin load balancing across all your rented GPUs. One request per GPU, in parallel.
Automatic failover
If a host goes offline, the pool routes around it. No errors, no downtime.
Model-aware routing
Request qwen3:32b and the pool finds the GPU running it. Different models on different GPUs, one endpoint.
One key, unlimited scale
Standard OpenAI SDK. Drop in your pool key. Works with Cursor, Cline, Aider, any OpenAI-compatible client.
from openai import OpenAI
# One endpoint. 10 GPUs behind it.
client = OpenAI(
base_url="https://gpusmarket.com/api/v1/openai/pool",
api_key="pool_abc123..."
)
# Every call hits a different GPU (round-robin)
for i in range(100):
r = client.chat.completions.create(
model="qwen3:32b",
messages=[{"role": "user", "content": f"Task {i}"}]
)Your GPU could be earning money right now
Thousands of GPUs sit idle while AI developers pay $12/hr to cloud providers for the same hardware. GPUsMarket connects them directly - no middleman, no markup. You earn, they save, AI becomes accessible to everyone.
Get started free